Product notes on hiring, not slide decks.

The problemA sourcing loop that keeps finding the same people

Most funnels are a Boolean search in a trench coat. Write the job description, pull the keywords, run the query, get back a list of people who used the right nouns. Volume was never the problem; signal is. Two candidates can match the same terms and be very different in the same role.

How it worksScore compatibility instead of running a query

Candidates see a Compatibility Index from 0 to 100 that compares a job's scouting profile (work style axes, salary, team size, work mode, perks, distance) with their own preferences. Recruiters see applicants ranked on skills, seniority and experience, salary, availability, location and language, with the breakdown per applicant.

• Skills scored on one 0–10 scale, shared by candidates and companies
• 70+ scouting-profile fields describe the organisation and the role
• Work style, salary, team size, work mode, perks and distance compared with a candidate's preferences
• A full match breakdown per applicant, not a single black-box number

In the productThe shortlist is the product

Recruiters open a ranked list of applicants with the breakdown per person: skills, seniority and experience, salary, availability, location and language. Candidates see their Compatibility Index on each job.

Why it mattersWhy structure matters

We have no outcome data to show yet, and we will not invent any. The argument is structural: when both the job and the candidate are described in structured data, the comparison is a calculation instead of a guess, and a recruiter can check why a name is on the list.

The problemScreening is the most repetitive work in hiring, and the first thing that gets rushed

The first three conversations with a candidate are largely the same every time: confirm the basics, probe the core skills, then see how the person works with others. At volume that is dozens of near-identical calls a week. They get rushed, skipped, or handed to whoever is free, and that inconsistency is exactly where strong candidates fall through and weak ones slip past.

How it worksConfigure the pipeline once; run it every time

Job posters set up the three-stage pipeline per role: what the pre-screening confirms, which competencies the technical round probes, and what the collaboration round looks for. The AI interviewer then runs that structure for every applicant, with the same questions and the same rubric, and scores each stage against the role.

• Pre-screening: confirms must-haves, availability and motivation before anyone spends a minute
• Technical interview: probes the role's core competencies against its skill rubric
• Collaboration interview: shows how the candidate reasons, communicates and handles ambiguity
• Configurable per role: which rounds run and what each one looks for
• Candidates interview on their own schedule; no calendar tetris for your team

In the productYour team reviews results, not a pile of CVs

Every applicant goes through the same rounds and arrives with a structured result: stage-by-stage scores, the quotes behind them, and a clear picture of strengths and risks. Recruiters open a ranked, pre-evaluated list instead of a hundred unscreened CVs.

Why it mattersWhy a fixed structure

A fixed structure makes the first conversations comparable. The same questions and the same rubric for everyone mean that differences between candidates come from the candidates, not from who happened to run the call. It also frees your team's time for the parts that need a human: judgment and candidate experience.

The problemAssistants with nothing structured to query

AI assistants that help with job search or sourcing mostly scrape public pages and parse free text, then guess. The assistant is not the weak link; the data source is. A structured, permissioned source gives it something to query instead of something to infer.

How it worksSixteen read-only tools, opt-in by the organisation and by the candidate

MirageScouting runs a remote MCP server with sixteen read-only tools. Organisations decide what is visible: publishing to agents is off by default, individual jobs can be hidden, only published stories and public events are shared, and hiring-team contact details are never returned. Candidates decide too: their profile is searchable only after they switch it on, it is off by default, and contact details are never part of it. Every result links back to its page on MirageScouting.

• search_jobs: find open roles with structured filters
• get_job: read one role in full
• search_companies: find organisations that opted in
• get_company_profile: read an opted-in organisation's published profile
• search_candidates: find candidates who opted in
• get_candidate_profile: read an opted-in candidate's restricted profile, with no contact details
• search_skills: look up the exact skill names used (no people data)
• search and fetch: one free-text search and a full read by id, for ChatGPT deep research
• Every data call is logged

In the productHow to connect

The server reads the same company, job and profile data as the rest of MirageScouting. Add it to Claude, Claude Code, ChatGPT, Gemini CLI or Gemini Enterprise with the install steps on the developer page; no MirageScouting account is needed to search.

Why it mattersWhy candidates stay in control

A profile is shared only when its owner opts in, and what is shared is limited to the professional profile: name, headline, location, experience, education, skill levels and similar. Switching the setting off removes the profile from search and its page stops resolving.

The problemA CV is a marketing document, not an evaluation

Candidates tune CVs for keyword density and recruiter attention. Recruiters scan for word presence, not proof. Neither side learns much about fit. The process reliably surfaces people who describe themselves well, which is not the same as people who can do the work.

How it worksOne scale, four bands, level descriptions

Skills are scored from 0 to 10 and grouped into four bands: Beginner (0–2), Intermediate (3–5), Advanced (6–8) and Master (9–10). Each skill has level descriptions, so a 7 means the same thing to a candidate and to a company. Interview results become verified levels only after review by the MirageScouting team.

• One 0–10 scale with four bands, from Beginner to Master
• Product management framework: 40 skills in 6 categories, each with level descriptions
• Strengths and gaps shown per skill and per category
• Interview results count as verified levels only after review by the MirageScouting team
• Frameworks for further roles are set up together with our team

In the productFrom assessment to sourcing filter

On the hiring side, recruiters can filter by skill and minimum level instead of keyword frequency. On the candidate side, people see where they stand and which gaps to close. The same skill scores feed the applicant ranking, so skills are not only visible but scored against the role.

Why it mattersWhy evidence over description

A CV is written to be read; a score is built to be compared. When the candidate and the role are both described in structured skill data, matching is a comparison rather than a guess, and a recruiter can see what a result rests on. We do not have outcome data to put a number on the difference yet.

The problemA certificate is a dead end; a profile should be alive

Most learning platforms hand you a PDF and stop. The gap that sent you there is closed, but nothing downstream knows it. Your profile still shows the old you, the next interview re-asks what you already proved, and any system reading your data is reading a snapshot that went stale the day it was written. Learning that doesn't travel is learning you have to keep re-proving.

How it worksOne confirmed skill level, shared by Academy and interview

MirageScouting keeps a single source of truth for what you can actually do. When the Academy confirms a skill, it doesn't write to its own private record; it updates the same confirmed skill level the interview pipeline reads. There is no second copy to drift out of sync. The level you proved by learning is the level the next interview sees, with an advance-only guard so a hard-won gain can't be silently downgraded.

• Confirmed skill levels live in one shared store: the Academy writes it, interviews read it
• Proof is verified, not self-claimed: your evidence is matched to the rubric before a level moves
• A learning summary is re-embedded after every milestone, so retrieval always sees the latest you
• Advance-only guard: a level you earned can't quietly slip backwards
• No second record to fall out of date: the profile and the proof are the same object

In the productFinish a module, and the loop closes on its own

Complete a module and three things happen without you lifting a finger. Your evidence is routed to the rubric and the confirmed skill level is updated. A fresh learning summary ("completed N courses across these skills, most recent first") is regenerated and re-embedded, replacing the previous one rather than stacking on top of it. From that moment your next interview can retrieve what you just learned.

Why it mattersWhy one record

Because the proof updates the profile in place and the summary is replaced at every milestone, your profile is never a snapshot; it is the current you. The freshness isn't a sync job someone has to remember to run. It is built into how learning is recorded.